Episode Summary
Executive Summary: This episode examines why big capital projects are slow, expensive, and prone to change orders, focusing on the misaligned incentives of traditional EPCs. Guest Alex Modon of Unlimited Industries argues that modern software and AI can let an EPC fully engineer projects upfront, reduce uncertainty, and offer true fixed-price delivery without change orders, starting with data centers and expanding across industrial infrastructure.
Main Topics: Why traditional EPC models fail (Priority: 5/5): The conversation centers on the structural mismatch between developers who want projects built faster and cheaper, and EPC contractors who profit from more time and higher costs under cost-plus or change-order-heavy fixed-firm contracts. Uncertainty in large capital projects (Priority: 5/5): Big projects are difficult to price because design, procurement, labor, and site conditions evolve over time. Traditional contracts are built to absorb this uncertainty, but that same structure weakens cost discipline. Engineering as the lever for cost certainty (Priority: 5/5): Modon argues that deeper front-end engineering can reduce uncertainty before financing and construction begin, allowing projects to reach a true issued-for-construction state before capital is committed. AI-enabled integrated engineering platform (Priority: 5/5): Unlimited Industries uses a unified software stack and AI to combine workflows that are usually split across CAD, spreadsheets, simulation tools, PDF review, and vendor management, speeding design and enabling broader optimization. Data centers as the initial beachhead (Priority: 4/5): Data centers are presented as a strong early use case because speed to power matters and requirements change rapidly, creating a large payoff for faster and more complete engineering. New business model: no change orders (Priority: 4/5): Unlimited aims to remove change orders entirely by taking on more upfront engineering and risk, relying on better definition, supplier visibility, and hedging to make fixed-price delivery actually credible. Broader implications for industrial construction (Priority: 4/5): The approach is positioned as applicable beyond data centers to power projects, battery storage, chemical plants, refineries, and other large infrastructure builds.
Key Arguments: Traditional EPC incentives are misaligned: contractors often make more money when projects cost more and take longer. Cost-plus contracts and tightly scoped fixed-firm contracts both encourage change orders rather than efficient delivery. The biggest source of improvement is not just engineering efficiency, but using engineering to reduce downstream procurement and construction risk. Most project development stops around 30% design completion before financing, leaving major uncertainty unresolved when commitments are made. Modern software can unify fragmented engineering workflows and let AI execute design trade studies using project-specific data and tools. A deeper design process creates better optimization, not just lower engineering labor costs. Data centers are especially attractive because design speed affects time to power, and their requirements are evolving rapidly. A true fixed-price model becomes feasible only if the contractor has far more upfront definition and vendor certainty than traditional EPCs have. Long-term vertical integration could further align incentives across engineering, subcontracting, procurement, and construction.
Data Points: Listeners survey incentive: $100 Amazon gift card - Mentioned in the opening promo for audience survey participation. Virtual power plant capacity: 3.4 gigawatts - EnergyHub ad claims its 2.5 million customer devices create dispatchable capacity equivalent to more than three nuclear reactors. Customer devices participating in VPPs: 2.5 million - EnergyHub ad describing device aggregation across thermostats, batteries, and EVs. Peak-period device shifts: Millions of thermostats, batteries, and EVs - EnergyHub ad notes these devices shifted energy during May and June peak periods. Typical project size: $100 million to many hundreds of millions, sometimes billions - Modon describes the scale range of large capital projects and why risk is substantial. Conceptual design estimate accuracy: Plus or minus 50% - Early-stage project cost estimate before front-end engineering. Front-end engineering estimate accuracy: Plus or minus 10% - Cost estimate after more detailed engineering, used for financing decisions. Engineering completion at financing: About 30% - He says projects often reach final investment decision before engineering is mostly complete. Remaining engineering after financing: About 70% - Large portion of design work still occurs after capital commitment. Engineering share of total project cost: 3% to 15% - Modon notes engineering is a small direct cost relative to total project cost. Design phase on data centers: 6 to 9 months - Current time budget for design that Unlimited aims to compress significantly. Target project timing: End of next year - Modon says he hopes to see the first project reach COD by then. Initial project delivery goal: No change orders - Unlimited plans to structure contracts without change orders. Vertical integration scope: Engineering, procurement, and construction - The company positions itself as a modern EPC with integrated workflows.
Pivotal Quotes: "As a third-party contractor, your incentive is really, we only make money when the project costs more and takes longer." — Shail Khan quoting the problem setup: Opening framing of the EPC incentive problem. "High-level, what's broken is it's the incentive structure." — Alex Modon: Core diagnosis of why traditional EPC delivery underperforms. "We literally put like no change orders in our contracts." — Alex Modon: Describing Unlimited Industries' fixed-price delivery model.
Implications: If AI-driven front-end engineering works as claimed, EPCs could shift from change-order businesses to true fixed-price builders, improving speed, cost certainty, and capital efficiency for data centers and industrial infrastructure.