Small modular reactors for data centers

Cell Reports Sustainability (2026), Perspective, co-first author

Nuclear & data centers
What would it take for small modular reactors to power data centers, and which economic dimensions must a credible techno-economic assessment capture?

Co-first author

  • Perspective article
  • Techno-economic section
  • FOAK and NOAK costs
  • Reliability premium
  • Levelized cost of computing

Research question

What would it take for small modular reactors to power data centers, and which economic dimensions must a credible techno-economic assessment capture?

Key insight

Reliability requirements add a capital premium that levelized cost leaves out

A hyperscale campus needs Tier IV-level availability, so one reactor is not enough: with N+1 redundancy, a 300 MW campus may need 400 MW of SMR capacity. Standard levelized-cost estimates omit this reliability premium, so they understate the cost of firm on-site supply.

My contribution

Co-first author. Wrote the techno-economic analysis section, introduction and conclusions.

How I approached it

  1. ChallengePublished SMR cost figures mix first-of-a-kind budgets with long-run nth-of-a-kind targets.

    What I didI set first-of-a-kind and nth-of-a-kind capital costs of SMR projects side by side and argued that assessments treat them as separate cases, each with its own financing and risk.

    Why it matteredLong-run figures make first projects look more viable than they are, and early data-center buyers carry that risk.

  2. ChallengeLevelized cost of electricity leaves out cooling, which takes 30–40% of a hyperscale facility's energy.

    What I didIn the section I drafted, I proposed a service-based metric, the levelized cost of computing, which puts electricity and cooling costs over units of computing service.

    Why it matteredIt compares on-site nuclear, grid and renewable-plus-storage supply by what each delivers to the data center.

Method

Perspective synthesizing technology, economics and policy for SMR-powered data centers.

Key findings

  1. Data-center reliability requirements add a substantial premium to the capital cost of on-site nuclear supply.
  2. First-of-a-kind and nth-of-a-kind cost estimates for SMRs diverge widely, so assumptions must be explicit.
  3. Service-based metrics, such as cost per unit of computing, and co-products such as heat matter for comparisons.

Boundary conditions

A Perspective: it frames the questions and evidence rather than presenting a new case-study model.

What it led to

Pricing the heat used for cooling, hour by hour, became the core of my boundary analysis for one data-center campus. Small modular reactor cogeneration for a hyperscale data center

Publications

  1. Cell Rep. Sustain.
    The role of small modular reactors in powering data centers for a sustainable digital economy
    R. A. Jacob*, H. Li*, A. M. Rad, B. Poudel, Y. Li, M. Sun, T. R. McJunkin, J. Zhang
    Cell Reports Sustainability, vol. 3, article 100820, 2026. Co-first authors
    DOI PDF