pith. sign in

arxiv: 2604.08854 · v1 · submitted 2026-04-10 · 📡 eess.SY · cs.GT· cs.SY

Risk-Aware Allocation of Transmission Capacity for AI Data Centers

Shaoze Li , Bohang Fang , Cong Chen This is my paper

Pith reviewed 2026-05-10 18:09 UTC · model grok-4.3

classification 📡 eess.SY cs.GTcs.SY
keywords risk-aware allocationtransmission capacityAI data centersfirm capacityflexible capacitysimultaneous ascending auctioncompetitive equilibriumrobust optimization
0
0 comments X

The pith

Tolerating minimal service interruption risk unlocks substantial flexible transmission capacity for AI data centers.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops robust optimization to set firm transmission capacity that minimizes unserved data center demand, then layers a risk-aware approach on top to release flexible capacity. It demonstrates that accepting a small probability of interruption and blackout can free up large amounts of additional grid capacity that would otherwise stay locked. The paper next applies a simultaneous ascending auction, with products defined by capacity amount, risk level, and location, to divide the scarce resources among competing data centers. A reader would care because the method could let new AI facilities connect faster without waiting for new lines to be built. The auction is shown to reach a competitive equilibrium and efficient outcome when valuations are additive or symmetric concave.

Core claim

By solving the robust optimization problem for firm capacity and then applying risk-aware allocation for flexible capacity, tolerating a minimal probability of service interruption and blackout unlocks substantial flexible capacity of transmission networks and accelerates data center interconnection. Under additive or symmetric concave valuation functions, the simultaneous ascending auction converges to a competitive equilibrium and achieves efficient allocation.

What carries the argument

Risk-aware allocation of flexible capacity, built on top of robust optimization for firm capacity, paired with a simultaneous ascending auction that characterizes products by capacity, risk level, and location.

If this is right

  • Transmission networks can interconnect more AI data centers without immediate new infrastructure.
  • The auction allocates scarce capacity efficiently among competing data centers.
  • Convergence to competitive equilibrium holds when valuations satisfy the stated conditions.
  • Accepting minimal blackout risk produces measurable gains in available flexible capacity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same split into firm and flexible capacity could apply to other variable loads such as electric vehicle charging stations.
  • Regulators might adopt similar risk-based auctions for allocating other constrained resources like interconnection queues.
  • Quantifying the exact capacity gain on a real transmission corridor would give a concrete test of the risk-tolerance benefit.
  • If the robust optimization step scales poorly on large networks, practical deployment would require new solvers or approximations.

Load-bearing premise

Data-center valuation functions are additive or symmetric concave so the auction converges, and the robust optimization problems for firm capacity can be solved efficiently.

What would settle it

Running numerical tests on realistic grid data where valuation functions violate additivity or symmetry and observing whether the auction fails to converge to equilibrium or produces inefficient outcomes would directly test the allocation result.

Figures

Figures reproduced from arXiv: 2604.08854 by Bohang Fang, Cong Chen, Shaoze Li.

Figure 1
Figure 1. Figure 1: A 4-bus transmission network example. While Proposition 2 has been previously established in [19] and [20], we provide an alternative, more direct proof in the online Appendix VI-E. Furthermore, Proposition 3 and Theorem 2 restate results established in [12]. IV. SIMULATIONS In this section, we present a case study detailing the bidding products from equations (1) and (4) in a transmission network. Data ce… view at source ↗
read the original abstract

Rapid growth in AI-driven data center loads is creating significant challenges for transmission grid interconnection. This paper proposes robust and risk-aware frameworks to quantify transmission capacity as firm and flexible capacities. We efficiently solve the robust optimization problem to determine firm capacity when minimizing unserved data center demand. Building upon this, we introduce a risk-aware allocation for flexible capacity, showing that tolerating a minimal probability of service interruption and blackout can unlock substantial flexible capacity of transmission networks and accelerate data center interconnection. To efficiently allocate scarce transmission capacities among competing data centers, we adopt the simultaneous ascending auction, characterizing products by capacity, risk level, and location. Under additive or symmetric concave valuation functions, the auction converges to a competitive equilibrium and achieves efficient allocation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes robust optimization frameworks to quantify firm transmission capacity for AI data centers by minimizing unserved demand, along with a risk-aware approach for flexible capacity that tolerates a minimal probability of service interruption and blackout to unlock additional network capacity. It further introduces a simultaneous ascending auction to allocate the resulting flexible capacity products (characterized by capacity, risk level, and location) among competing data centers, claiming that the auction converges to a competitive equilibrium and achieves efficient allocation when data-center valuation functions are additive or symmetric concave.

Significance. If the robust optimization is computationally tractable and the auction convergence holds under realistic conditions, the work could meaningfully accelerate data-center interconnections by enabling greater utilization of existing transmission assets with controlled risk. The market-based allocation via auction is a positive feature for handling competing loads. However, the assessed significance is tempered by the absence of derivations, numerical validation, or error analysis in the available text, which prevents verification of the claimed capacity gains or equilibrium properties.

major comments (2)
  1. [Abstract] Abstract: The central claim that the simultaneous ascending auction converges to a competitive equilibrium and achieves efficient allocation is stated conditionally on additive or symmetric concave valuation functions, but the manuscript provides no derivation, proof sketch, or specific reference to the auction-theory result invoked, nor any mapping of typical AI data-center valuations (which may include complementarities across locations or risk levels) to these assumptions.
  2. [Abstract] Abstract and main text: The robust optimization problem for firm capacity is described as efficiently solvable when minimizing unserved demand, yet no algorithm, complexity bound, or numerical example is supplied; without these, the claim that substantial flexible capacity can be unlocked cannot be assessed for practicality or sensitivity to the minimal interruption probability parameter.
minor comments (1)
  1. [Abstract] The characterization of auction products by capacity, risk level, and location is introduced without explicit mathematical notation or example definitions, which could be clarified for readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments correctly identify areas where the manuscript would benefit from greater explicitness on theoretical foundations and computational details. We address each point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that the simultaneous ascending auction converges to a competitive equilibrium and achieves efficient allocation is stated conditionally on additive or symmetric concave valuation functions, but the manuscript provides no derivation, proof sketch, or specific reference to the auction-theory result invoked, nor any mapping of typical AI data-center valuations (which may include complementarities across locations or risk levels) to these assumptions.

    Authors: The observation is accurate: the abstract states the convergence result without a proof sketch or explicit reference. The manuscript relies on standard results from the auction-theory literature on simultaneous ascending auctions for multi-item settings, which establish convergence to competitive equilibrium under additive valuations and under symmetric concave valuations. We will add a concise reference to the relevant result (e.g., from the literature on SAA convergence for concave valuations) together with a short explanatory paragraph in the revised abstract and Section on market allocation. We will also include a brief discussion of valuation assumptions, noting that while real AI data-center preferences may exhibit complementarities across locations or risk levels, the efficiency guarantees hold under the stated classes and the framework can be extended or used as a benchmark. revision: yes

  2. Referee: [Abstract] Abstract and main text: The robust optimization problem for firm capacity is described as efficiently solvable when minimizing unserved demand, yet no algorithm, complexity bound, or numerical example is supplied; without these, the claim that substantial flexible capacity can be unlocked cannot be assessed for practicality or sensitivity to the minimal interruption probability parameter.

    Authors: We acknowledge that the current text does not supply an explicit algorithm, complexity statement, or numerical illustration for the robust optimization. The formulation is a robust linear program over an uncertainty set on demand and line capacities; it is solved via standard LP solvers after dualization or scenario-based approximation. In the revision we will insert a short description of the solution procedure, a polynomial-time complexity bound for fixed uncertainty-set dimension, and a minimal numerical example (e.g., on a small test network) that shows the additional flexible capacity obtained as a function of the tolerated interruption probability. This will allow readers to assess practicality and sensitivity. revision: yes

Circularity Check

0 steps flagged

No circularity; claims rest on external auction-theory results under stated assumptions

full rationale

The paper quantifies firm capacity via robust optimization and allocates flexible capacity via simultaneous ascending auction, stating convergence to competitive equilibrium only under the explicit external condition that valuation functions are additive or symmetric concave. No equations, parameters, or results are fitted inside the paper and then renamed as predictions; no self-citations are invoked as load-bearing uniqueness theorems; the derivation chain applies known methods to the data-center setting without reducing outputs to inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Central claims rest on domain assumptions from optimization and mechanism design plus one tunable risk parameter; no new physical entities or fitted constants beyond the risk tolerance level are introduced.

free parameters (1)
  • minimal interruption probability
    Chosen tolerance level that determines how much extra flexible capacity is released; appears as a design parameter in the risk-aware allocation step.
axioms (1)
  • domain assumption Data-center valuation functions are additive or symmetric concave
    Invoked to guarantee convergence of the simultaneous ascending auction to competitive equilibrium.

pith-pipeline@v0.9.0 · 5420 in / 1138 out tokens · 57902 ms · 2026-05-10T18:09:45.009252+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    AI, data centers, and the U.S. electric grid: A watershed moment,

    Belfer Center for Science and International Affairs, “AI, data centers, and the U.S. electric grid: A watershed moment,” https://www.belfercenter.org/research-analysis/ai-data-centers-us- electric-grid, 2025

    work page 2025

  2. [2]

    2024 United States data center energy usage report,

    A. Shehabi, S. J. Smith, M. Huber, E. Masanet, and D. Sartor, “2024 United States data center energy usage report,” Lawrence Berkeley National Laboratory, Berkeley, California, Tech. Rep. LBNL-2001637, 2024

    work page 2024

  3. [3]

    Order on co-location of large loads at generating facilities in PJM,

    Federal Energy Regulatory Commission, “Order on co-location of large loads at generating facilities in PJM,” 2025

    work page 2025

  4. [4]

    The role of flexible connection in accelerating load interconnection in distribution networks,

    N. Gu, G. Chen, and J. Qin, “The role of flexible connection in accelerating load interconnection in distribution networks,”arXiv preprint arXiv:2510.11476, 2025

    work page arXiv 2025

  5. [5]

    A new milestone for smart, affordable electric- ity growth,

    M. Terrell, “A new milestone for smart, affordable electric- ity growth,” https://blog.google/innovation-and-ai/infrastructure-and- cloud/global-network/demand-response-data-center-milestone/, 2026

    work page 2026

  6. [6]

    Energy management for renewable-colocated artificial intelligence data centers,

    S. Li, L. Tong, and T. D. Mount, “Energy management for renewable-colocated artificial intelligence data centers,”arXiv preprint arXiv:2507.08011, 2025

    work page arXiv 2025

  7. [7]

    Optimization-based hosting capacity enhancement with device coordination,

    A. Almutairiet al., “Optimization-based hosting capacity enhancement with device coordination,”Electric Power Systems Research, 2024

    work page 2024

  8. [8]

    Hosting capacity enhancement through coordinated optimization,

    M. Kamruzzaman and M. Benidris, “Hosting capacity enhancement through coordinated optimization,”IEEE Transactions on Industry Applications, 2020

    work page 2020

  9. [9]

    Rethinking load growth: Assessing the potential for integration of large flexible loads in U.S. power systems,

    T. H. Norris, T. Profeta, D. Pati ˜no-Echeverri, and A. Cowie-Haskell, “Rethinking load growth: Assessing the potential for integration of large flexible loads in U.S. power systems,” Nicholas Institute for Energy, Environment & Sustainability, Duke University, Tech. Rep., 2025

    work page 2025

  10. [10]

    International principles for the prioritisation of grid connections,

    T. C. Group, “International principles for the prioritisation of grid connections,” Energimarknadsinspektionen, Tech. Rep., 2020

    work page 2020

  11. [11]

    An assessment of large load interconnection risks in the western interconnection,

    R. Quint, J. J. Zhao, and K. Thomas, “An assessment of large load interconnection risks in the western interconnection,” Western Electricity Coordinating Council, Tech. Rep., 2025

    work page 2025

  12. [12]

    Putting auction theory to work: The simultaneous as- cending auction,

    P. Milgrom, “Putting auction theory to work: The simultaneous as- cending auction,”Journal of Political Economy, vol. 108, no. 2, pp. 245–272, 2000

    work page 2000

  13. [13]

    The FCC spectrum auctions: An early assessment,

    P. Cramton, “The FCC spectrum auctions: An early assessment,” Journal of Economics & Management Strategy, vol. 6, no. 3, pp. 431– 495, 1997

    work page 1997

  14. [14]

    Spectrum auctions,

    ——, “Spectrum auctions,”Handbook of Telecommunications Eco- nomics, vol. 1, pp. 605–639, 2002

    work page 2002

  15. [15]

    Power system analysis: a mathematical approach,

    S. Low, “Power system analysis: a mathematical approach,”Lecture Notes for EE/CS/EST, vol. 135, 2022

    work page 2022

  16. [16]

    Wholesale market partic- ipation of DERAs:DSO-DERA-ISO coordination,

    C. Chen, S. Bose, T. D. Mount, and L. Tong, “Wholesale market partic- ipation of DERAs:DSO-DERA-ISO coordination,”IEEE Transactions on Power Systems, vol. 39, no. 5, pp. 6605–6614, 2024

    work page 2024

  17. [17]

    Auction of flexible-use service licenses in the 3.45–3.55 GHz band for next-generation wireless services,

    F. C. Commission, “Auction of flexible-use service licenses in the 3.45–3.55 GHz band for next-generation wireless services,” Federal Communications Commission, Washington, DC, Tech. Rep. DA 21- 655, 2021

    work page 2021

  18. [18]

    Risk-based capacity accreditation of resource-colocated large loads in capacity markets,

    S. Li, L. Tong, and T. D. Mount, “Risk-based capacity accreditation of resource-colocated large loads in capacity markets,”arXiv preprint arXiv:2511.17715, 2025

    work page arXiv 2025

  19. [19]

    Competitive equilibrium in an exchange economy with indivisibilities,

    S. Bikhchandani and J. W. Mamer, “Competitive equilibrium in an exchange economy with indivisibilities,”Journal of Economic Theory, vol. 74, no. 2, pp. 385–413, 1997

    work page 1997

  20. [20]

    A note on Kelso and Crawford’s gross substitutes condition,

    S. Fujishige and Z. Yang, “A note on Kelso and Crawford’s gross substitutes condition,”Mathematics of Operations Research, vol. 28, no. 3, pp. 463–469, 2003

    work page 2003

  21. [21]

    Conditional value-at-risk for general loss distributions,

    R. T. Rockafellar and S. Uryasev, “Conditional value-at-risk for general loss distributions,”Journal of Banking & Finance, vol. 26, no. 7, pp. 1443–1471, 2002. ACKNOWLEDGMENT The authors would like to thank Prof. Junjie Qin from Purdue University and Dr. Zhenyu Fan from PJM for their insightful comments, which significantly improved the qual- ity of this ...

    work page 2002

  22. [22]

    Furthermore, there existsε >0such thatc+εe j ∈ P I , wheree j is thej-th standard basis vector

    lettingE j denote the set of edges on the unique path from the root to nodej, one hasE j ∩ T(c) =∅. Furthermore, there existsε >0such thatc+εe j ∈ P I , wheree j is thej-th standard basis vector. Proof of Lemma 1.For each edgee, letD e ⊆ {1, . . . , N} denote the set of downstream nodes ofe. The entries ofA are defined asA e,i = 1ifi∈ D e, and0otherwise, ...

    work page

  23. [23]

    The modified surplus for the final allocation is ˆV1({3,4})−p ∗({3,4}) = 30.We compare this against all other relevant subsets in Table VI

    Example 1:For bidder 1, we verify thatU ∗ 1 maximizes the modified surplus ˆV1(U)−p ∗(U). The modified surplus for the final allocation is ˆV1({3,4})−p ∗({3,4}) = 30.We compare this against all other relevant subsets in Table VI. Since Bidder 1 has zero valuation for Item 2, any bundle containing Item 2 is dominated and thus omitted. TABLE VI MODIFIED SUR...

    work page

  24. [24]

    For Bidder 1, the maximum modified net utility is 30, achieved uniquely atU ∗ 1 ={2,4}

    Example 2:Table VIII and IX report these surpluses for each bidder. For Bidder 1, the maximum modified net utility is 30, achieved uniquely atU ∗ 1 ={2,4}. Hence, Con- dition (7a) holds for Bidder 1. For Bidder 2, the maximum modified net utility is 25, achieved atU ∗ 2 ={1,3}as well as at{1,2,3}and{1,3,4}. SinceU ∗ 2 is among the maximizers, Condition (7...

    work page